Heart rate variability in relation to prognosis after myocardial infarction: Selection of optimal processing techniques

Automatic analysis of heart rate variability from Holter recordings may be invalidated by beat recognition errors and recording artefact, necessitating filtering and editing of the computer-recognized RR interval sequence. Two new methods for heart rate variability analysis have been developed, based on an estimation of the width of the main peak of the frequency distribution curve of the computer-recognized normal-to-normal beat sequence. These methods are independent of a low level of recognition error and artefact, thus removing the need for operator-dependent, time-consuming editing. The value of the new methods (heart variability indices 1 and 2) in identifying patients with serious events (death and symptomatic, sustained documented ventricular tachycardia) during a 6-month follow-up after acute myocardial infarction was assessed in a case-control study comparing 20 patients who had experienced such events (Group I) with 20 patients who, following admission with acute myocardial infarction, had remained free of complications for greater than 6 months after discharge (Group II). Group II was selected to match Group I with regard to age, sex, infarct site, ejection fraction, and beta-blocker treatment. Analysis of the unfiltered computer-recognized normal-to-normal interval sequence showed that heart rate variability indices 1 and 2 were significantly lower (P less than 0.005, P less than 0.002) in those who had experienced events compared with those free from complications. Two other methods of expressing heart rate variability, including the standard deviation method, in combination with four different data-filtering techniques, gave less significant distinction between those with and without events during follow-up. It is concluded that using the methods described, reduced heart rate variability in patients at risk from death or sustained ventricular tachycardia after acute myocardial infarction can be detected automatically from unfiltered Holter tape recordings even in the presence of a low level of beat recognition error and recording artefact.